Extend GenAI system to support IBM Watsonx AI and AWS Bedrock

[gyliu513]

Added IBM Watsonx AI and AWS Bedrock to the list of possible values that the gen_ai.system attribute can take.

Changes

The following are added:

• IBM Watsonx AI
• AWS Bedrock

Please provide a brief description of the changes here.

Note: if the PR is touching an area that is not listed in the existing areas, or the area does not have sufficient domain experts coverage, the PR might be tagged as experts needed and move slowly until experts are identified.

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• Change log entry added, according to the guidelines in When to add a changelog entry.
  • If your PR does not need a change log, start the PR title with [chore]
• schema-next.yaml updated with changes to existing conventions.

[gyliu513]

@lmolkova @lzchen can you help review? I want to start the process of contributing watsonx to genai instrumentation, thanks!

[lmolkova]

Thank you!

[lmolkova]

@open-telemetry/semconv-genai-approvers please take a look

[drewby]

Looks good to me.

[gyliu513]

@lmolkova do you know when this will be reflected to https://github.com/open-telemetry/opentelemetry-python/blob/main/opentelemetry-semantic-conventions/src/opentelemetry/semconv/_incubating/attributes/gen_ai_attributes.py ? Thanks

[lmolkova]

@gyliu513 we'll need to release the new version of semconvs first. We're due for release, but with KubeCon we got behind and it might take 1-2 weeks.

Then someone (usually me, but can be anyone else) needs to regenerate code in https://github.com/open-telemetry/opentelemetry-python/tree/main/opentelemetry-semantic-conventions.

[mxiamxia]

Hi @gyliu513 @lmolkova , is there a convention for specifying GenAI attribute values or, more generally, in OpenTelemetry? For instance, _ is used as delimiter in AWS service attribute values, like aws_sqs in Java SDK below. And this proposal uses aws.bedrock. Is it more preferred way to use . delimiter for gen_ai.system attribute specifically?

https://github.com/open-telemetry/opentelemetry-java-instrumentation/blob/c6a6fa622465ebf931ed5860987cb0ecc6a04a16/instrumentation/aws-lambda/aws-lambda-events-2.2/library/src/main/java/io/opentelemetry/instrumentation/awslambdaevents/v2_2/internal/SqsEventAttributesExtractor.java#L23

[gyliu513]

@mxiamxia I think there is no standard for this yet, you may also see that vertex ai is using vertex_ai. Do you want to change the value to aws_bedrock?

@lmolkova I think it will be great if we can keep consistent for all LLM platforms, like what is the recommended delimiter for the attribute values.

[lmolkova]

. should be used by default. _ should be used to separate words that make no sense without each other - https://github.com/open-telemetry/semantic-conventions/blob/main/docs/general/attribute-naming.md

Think about it as namespaces in the code. There could be a lot of different things in aws namespace, so aws.bedrock, but it'd be rate_limiting instead of rate.limiting.

It took us a bit to come up with this rule of thumb, so you might see a lot of examples in the repo that don't follow this patter yet (or not worth breaking). Anyway, we try to follow this pattern for everything new.

[mxiamxia]

Thank you all for the response. AWS Observability team is working on supporting AWS Bedrock in CloudWatch by adhering to the OpenTelemetry LLM Spec. We will follow the aligned pattern and collaborate with you to contribute the relevant changes in OTel SDKs.